Implicit conversion from float -18.611111111111114 to int loses precision